Thanks for the review. Are you OK with un-statitcize
DiagnoseAccessPath so that SemaDeclCXX can call it -- or do you prefer
to introduce a wrapper?
On Sun, Jul 26, 2015 at 5:01 PM, Richard Smith <rich...@metafoo.co.uk> wrote:
> rsmith added inline comments.
>
> ================
> Comment at: lib/Sema/SemaDeclCXX.cpp:5575-5581
> @@ +5574,9 @@
> + // Explain the reasons of inaccessibility.
> + if (Field) {
> + if (CXXRecordDecl *C = Field->getType()->getAsCXXRecordDecl()) {
> + CXXDestructorDecl *D = S.LookupDestructor(C);
> + if (D && !D->isImplicit())
> + S.Diag(D->getLocStart(), diag::note_destructor_here);
> + }
> + }
> + }
> ----------------
> This doesn't seem right: you're producing a note for a destructor no matter
> which special member function was found to be inaccessible. The best thing to
> do would be to ask the SemaAccess code to produce the relevant note (see
> DiagnoseAccessPath); that way you can produce a note that says "declared
> private here" or whatever else is relevant (which can sometimes be that a
> virtual base destructor is inaccessible due to a private access path, for
> instance).
>
>
> Repository:
> rL LLVM
>
> http://reviews.llvm.org/D11517
>
>
>
_______________________________________________
cfe-commits mailing list
cfe-commits@cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
